BEGIN:VCALENDAR
VERSION:2.0
CALSCALE:GREGORIAN
PRODID:iCalendar-Ruby
BEGIN:VEVENT
CATEGORIES:Lecture / Talk / Workshop
DESCRIPTION:Anderson Ye Zhang\nWharton School of the University of Pennsylv
ania\n\nAbstract: Ranking from pairwise comparisons is a central problem in
a wide range of learning and social contexts. Researchers in various disci
plines have made significant methodological and theoretical contributions t
o it. However\, many fundamental statistical properties remain unclear espe
cially for the recovery of ranking structure. This talk presents two recent
projects towards optimal ranking recovery\, under the Bradley-Terry-Luce (
BTL) model.\nIn the first project\, we study the problem of top-k ranking.
That is\, to optimally identify the set of top-k players. We derive the min
imax rate and show that it can be achieved by MLE. On the other hand\, we s
how another popular algorithm\, the spectral method\, is in general subopti
mal.\nIn the second project\, we study the problem of full ranking among al
l players. The minimax rate exhibits a transition between an exponential ra
te and a polynomial rate depending on the magnitude of the signal-to-noise
ratio of the problem. To the best of our knowledge\, this phenomenon is uni
que to full ranking and has not been seen in any other statistical estimati
on problem. A divide-and-conquer ranking algorithm is proposed to achieve t
he minimax rate.
DTEND:20210227T000000Z
DTSTAMP:20211201T145752Z
DTSTART:20210226T230000Z
LOCATION:
SEQUENCE:0
SUMMARY:Probability and Statistics Seminar: Optimal Ranking Recovery from P
airwise Comparisons
UID:tag:localist.com\,2008:EventInstance_35971404826287
URL:https://calendar.usc.edu/event/probability_and_statistics_seminar_optim
al_ranking_recovery_from_pairwise_comparisons
END:VEVENT
END:VCALENDAR